Mathematical models for analysing the immune response to influenza vaccination in a patient population
Open access
Author
Date
2021Type
- Doctoral Thesis
ETH Bibliography
yes
Altmetrics
Abstract
Influenza viruses cause respiratory infections and spread in yearly outbreaks worldwide, causing up to 650 thousand deaths every year. Vaccination can induce protective antibodies against influenza viruses and is the most successful strategy to prevent influenza infections. Unfortunately, influenza viruses rapidly evolve and escape immunity established in previous vaccinations. Frequent updates of the influenza vaccine formulation and continuous evaluation of vaccine efficacy in large populations are necessary. A detailed molecular characterization of vaccine responses is experimentally very difficult and often unfeasible in larger populations. Instead, the hemagglutination inhibition (HI) antibody titer is commonly used as an easily accessible measure for the potency of influenza vaccine responses.
In this thesis, we present three mathematical models for analysing HI titers and propose how they can be used in conjunction to characterize the vaccine response in a patient population based on easily accessible measurements and medical record information. Specifically, we apply the models to patients after hematopoietic stem cell transplantation (HSCT), a high-risk group eliciting heterogeneous vaccine responses that are not well understood.
We first identify patient factors associated with HI titers for three different influenza strains. We show that sequential regression models are superior to the commonly used binary regression on conventional cut-offs (seroconversion/seroprotection) for inferring associations between patient factors and HI titers. Both approaches have a similar interpretation and yield consistent results, but sequential regression models infer associations with higher precision.
Next, we present a biophysical model of the HI assay that establishes a quantitative relationship between antibody concentration, antibody avidity (binding strength) and HI titer. We apply the model to infer antibody avidities in HSCT patients from antibody concentrations and HI titers, and experimentally validate our predictions. The model predicts that influenza vaccination mostly induced an increase in antibody concentration but not in avidity. Because the model links antibody concentrations and avidities to HI titers, it enables to connect mathematical models of the immune response to HI titers assessed in vaccine studies.
Finally, we integrate the identified most important patient factors into a dynamic model of the vaccine response in HSCT patients and combine it with the biophysical measurement model of the HI assay to infer patient-specific differences in immune response mechanisms. Specifically, we infer differences in memory B cells and germinal center (GC) processes that are potentially modulated by the investigated patient factors. The model predicts that vaccination induced antibody production by both plasma B cells from GCs and reactivated memory B cells. The heterogeneity in HI titer responses was well described by memory B cells and only a few patient factors that potentially affect the GC response (lymphocyte count and IFN-lambda genotype). The study demonstrates how dynamic modelling of the immune response can be combined with clinical patient information and statistical inference to characterize the vaccine response based on HI titers. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000504058Publication status
publishedExternal links
Search print copy at ETH Library
Contributors
Examiner: Stelling, Jörg
Examiner: Egli, Adrian
Examiner: Höfer, Thomas
Examiner: Maathuis, Marloes H.
Publisher
ETH ZurichSubject
influenza vaccine response; Hemagglutination inhibition assay; Antibody response; Antibody avidity; Germinal center (GC) B cells; Hematopoietic stem cell transplantation; dynamic modelling; categorical regressionOrganisational unit
03699 - Stelling, Jörg / Stelling, Jörg
More
Show all metadata
ETH Bibliography
yes
Altmetrics